Mineralogy and Petrology (2020) 114:453–463 https://doi.org/10.1007/s00710-020-00709-9
ORIGINAL PAPER
Mid-infrared spectroscopy of alkali feldspar samples for space application
Maximilian P. Reitze1 · Iris Weber1 · Herbert Kroll2 · Andreas Morlok1 · Harald Hiesinger1 · Jorn¨ Helbert3
Received: 18 November 2019 / Accepted: 18 May 2020 / Published online: 30 July 2020 © The Author(s) 2020
Abstract Feldspars are major components of terrestrial planetary surfaces. For future space application and the setup of a comprehensive reference database, Na- and K-rich alkali feldspars, NaAlSi3O8 – KAlSi3O8, have been investigated by infrared reflectance spectroscopy. We related the feldspar spectra to the chemical composition and state of Al,Si order/disorder. The infrared measurements were analyzed with respect to band shifts and peak shapes using the autocorrelation function. Natural samples served as starting materials. Some samples were treated by the alkali exchange method to produce pure end-members, which were then heated to generate various states of Al,Si disorder. X-ray diffraction (XRD) methods served to determine the Al,Si distribution. Our autocorrelation allowed to differentiate between the compositional and the order/disorder influences seen in the spectra in the wavelength range between 7 μmupto14μm (1429 cm−1 to 714 cm−1). Space missions often analyze the surfaces of planetary bodies using remote sensing. Therefore, our results are essential to characterize and distinguish alkali feldspars on the surfaces of terrestrial planetary bodies like Mercury.
Keywords Alkali feldspars · Infrared spectroscopy · X-ray diffraction · Al,Si distribution
Introduction feature (CF) at 8 μm (1250 cm−1) near the equator at approximately 120◦ longitude. After spectral unmixing of Feldspars are rock-forming minerals, on Earth and on Earth-based observations, Mercury’s surface is thought to other terrestrial planetary bodies. In the Earth’s crust, they contain potassium-rich feldspar (Sprague et al. 2009). account for more than 50 vol% (Okrusch and Matthes Direct sampling of Solar System bodies other than Earth 2014). Lunar highlands, which cover approximately 83% requires landing on the surface and returning samples, of the lunar surface (Jolliff et al. 2006), are composed which is extremely expensive and currently not feasible mostly of anorthosites and gabbros with a large amount of for most objects. Thus, remote sensing data of planets and anorthite (Wood et al. 1970). Furthermore, Mercury’s crust moons are necessary to provide the missing information. is feldspatic with a labradorite-like composition predicated Between 2011 and 2015, the MESSENGER (MErcury on the basis of Earth-based infrared emission spectroscopy Surface, Space ENvironment, GEochemistry and Ranging) by Sprague et al. (1997), who measured a Christiansen mission investigated the hermean surface (see Solomon et al. 2018, for a detailed description). The gamma-ray spectrometer (GRS) onboard the MESSENGER mission Editorial handling: A. Beran measured high Mg/Si and low Al/Si and Ca/Si contents, Maximilian P. Reitze which excludes an anorthite-rich feldspatic crust like the [email protected] lunar highlands (Nittler et al. 2011). GRS detected relatively high abundances of sodium, which suggest that albite is a 1 Institut f¨ur Planetologie, Westfalische¨ Wilhelms-Universitat,¨ major mineral on the surface of Mercury (Evans et al. 2012). Wilhelm-Klemm-Str. 10, 48149, M¨unster, Germany In October 2018, the ESA/JAXA BepiColombo space- 2 Institut f¨ur Mineralogie, Westfalische¨ Wilhelms-Universitat,¨ craft was launched to investigate the innermost planet of Corrensstr. 24, 48149, M¨unster, Germany our Solar System via remote sensing. Among several other 3 Deutsches Zentrum f¨ur Luft- und Raumfahrt, Rutherfordstr. 2, instruments, MERTIS (MErcury Radiometer and Thermal 12489, Berlin, Germany Infrared Spectrometer) will investigate Mercury’s surface 454 M.P. Reitze et al. in the mid-infrared (mid-IR) wavelength region (Hiesinger modification of microcline that under the polarizing micro- et al. 2010; H. Hiesinger and co-workers, in preparation). scope shows a characteristic tartan twinning reflecting the The goals of MERTIS are to study the surface mineralogy loss of monoclinic symmetry elements (Parsons 1978). The and composition, to map the mineralogy, and to measure the phase transition causes the T1 and T2 sites to split into T10, surface temperatures as well as the thermal inertia (Hiesinger T1mandT20, T2m sites, respectively. Aluminum concen- et al. 2010; H. Hiesinger and co-workers, in preparation). trates in the T10 site so that at full order t10 = 1, t1m = To achieve these goals, MERTIS operates in the wavelength t20 = t2m = 0. At less favorable conditions, due to the region from 7 μmto14μm (Spectrometer Channels/TIS high kinetic Al,Si exchange barrier, the transition results in a 1429 cm−1 to 714 cm−1) and from 7 μmto40μm (Radio- kinetically stranded, metastable pseudomorph called ortho- meter Channels/TIR 1429 cm−1 to 250 cm−1) (Hiesinger clase. On a submicroscopic scale, orthoclase consists of et al. 2010; H. Hiesinger and co-workers, in preparation). intimately intergrown triclinic domains. All textural stages Reflectance and absorption are not directly convertible between orthoclase and microcline may be present in one (e.g., Kirkland et al. 2002; Thomson and Salisbury 1993). and the same morphological unit. Na-rich alkali feldspars Therefore to apply spectral deconvolution algorithms we are as well symmetrically monoclinic at high temperature need reflectance spectra, which are comparable to remotely (monalbite). Changes to triclinic symmetry are induced acquired data for the remote sensing applications. during cooling through two different, though coupled mech- ◦ For the interpretation of data returned from space anisms: A displacive transition at 980 C, followed by a missions, laboratory work is essential. Known factors which continuous Al,Si ordering process (Salje et al. 1985). On ◦ control reflectance and emission mid-IR data are, for slow cooling, Al,Si ordering proceeds at T < 980 Cin example, the chemical composition, grain size distribution, the triclinic state with Al gradually enriched in the T10 site bulk sample density, viewing geometry, pressure, and while leaving the other three sites so that in the end t10 = 1, temperature (e.g., Pieters and Englert 1993). In this context, t1m = t20 = t2m = 0 like in microcline. By contrast, when analyses of different grain sizes are necessary, because the rapidly cooled, monalbite collapses to triclinic metastable surfaces of atmosphereless planetary bodies, like Mercury, analbite because the Na ion is too small to prevent structural are highly affected by larger and smaller impacts, which collapse in contrast to the larger K ion in sanidine. The terms cause the formation of a poorly-sorted regolith. high albite and low albite denote different states of order. At To setup our spectral database that is valuable for remote high temperatures, the solid solution series between monal- sensing of various Solar System bodies, it is necessary to bite and sanidine is continuous. A wide miscibility gap opens investigate well characterized samples of the alkali feldspar at lower temperatures, leading to the formation of intimately system. This includes both chemical composition and degree intergrown phase segregation textures of which perthites of Al,Si order in addition to other known factors controlling (Na-feldspar in a matrix of K-feldspar) are most common the shape of mid-IR spectra (Weber et al. 2018). (Okrusch and Matthes 2014, and referenes therein). The study of feldspars includes their formation con- Characteristics of alkali feldspars concerning the struc- ditions, occurrences, chemical, and physical behaviors. ture, formation conditions, exsolution processes, and behav- Alkali feldspars are solid solutions of the end-members ior at high temperatures are well known (e.g., Ribbe 1983). K[AlSi3O8] (Or) and Na[AlSi3O8] (Ab). They mostly con- Previous infrared spectroscopy work on the feldspar system tain small amounts of Ca[Al2Si2O8] (An). In addition to shows that changes in absorption spectra occur according their chemical compositions, they are characterized by the to chemical composition and degree of Al,Si order (e.g., specific distributions of Al and Si onto the tetrahedral sites Laves and Hafner 1956; Hafner and Laves 1957; Martin of their framework structures (“structural state”). The most 1970; Zhang et al. 1996, 1997). Hecker et al. (2010)gave common rock-forming alkali feldspars, ranging in compo- a review of different infrared techniques and their use- sition from Or100Ab0 to Or30Ab70, occur in two different fulness for remote sensing application. Iiishi et al. (1971) modifications. Monoclinic sanidine (K, Na)[AlSi3O8] is the correlated single bands in K-feldspar absorption spectra stable modification at high temperatures. The Al and Si between 8.7 μm and 23.3 μm (1150 cm−1 and 430 cm−1) atoms (1 Al plus 3 Si per formula unit) are distributed in a with specific stretching, bending, translation, and expan- highly disordered, but not completely random manner onto sion/contraction vibrations of Si-O, Si-Si, and Si-O-Si two T1 and two T2 tetrahedral sites. The larger Al atom bonds, which are controlled by the state of Al,Si order. How- slightly prefers the T1 over the T2 site (Scambos et al. 1987), ever, most of the bands described by Iiishi et al. (1971) denoted by t1 > t2 with t = Al/(Al + Si) of the respec- cannot be observed as single peaks in powder reflectance tive site (Kroll 1973). Provided a rate-enhancing influence spectra. being present, such as a fluid phase, mineral-fluid replace- The objective of our study is to measure mid-IR powder ment processes lead at T ≤ 450 ◦C - 480 ◦C to the triclinic reflectance spectra of well characterized alkali feldspar Mid-infrared spectroscopy of alkali feldspar samples for space application 455 samples to build a database for remote sensing applications. 1 This database is necessary especially for MERTIS. t
Samples and analyses Ab/Or
Samples
To differentiate and quantify controlling factors of the powder mid-IR reflectance spectra, the analyzed samples must be well characterized. Natural alkali feldspars were
used as raw materials, which were first thoroughly γV characterized and then treated in different ways described in the subsection “Sample Treatment”. All samples presented in this study, both natural and treated, have an internal
database identification number (ID, Table 1), which refers β to our database which is accessible at www.uni-muenster. de/Planetology (Weber et al. 2018). Our starting material included sanidine from Volkesfeld, Eifel, Germany (ID 73), microcline from Prilep, North Macedonia (ID 151), α and albite from Cazadero, California (ID 125). ID 73 sanidine, (Na15K85)[AlSi3O8], is of gem-stone quality, being transparent. Its Al,Si order is low, 2t1 = 0.61 (Kroll and Knitter 1991). ID 151 microcline, (Na5K95)[AlSi3O8], is a light greenish powder sample fraction (100 μmto 125 μm) prepared from the original perthitic microcline (amazonite) from Prilep, which contains some exsolved albite. The Al content of the T10 site is 0.99 (Strob 1983). ID 125 albite has a white color and represents an endmember of the alkali feldspar solid solution in terms of Ab content (Ab99.9) and Al,Si order (t10 = 1.00) (Wenk and Kroll 1984). These natural samples were treated in different ways as described in section sample treatment. O/ 18dO/ 8.142(4) 47d 8.137(3) 12.846(6) 12.826(6) 7.126(3) 7.130(2) 93.92(5) 116.47(4) 94.02(4) 116.47(2) 89.33(5) 665.5(5) 89.08(4) 664.4(4) 100/0 100/0 0.79(4) 0.77(5) 2 2 O/ 8dO/ 15d 8.146(3) 8.144(4) 12.799(4) 12.816(8) 7.144(2) 7.149(2) 94.22(3) 94.21(5) 116.57(2) 116.61(3) 88.07(3) 87.93(5) 664.4(3) 665.3(6) 100/0 100/0 0.90(3) 0.86(7) O/ 92d 8.158(2) 12.874(3) 7.111(1) 93.51(2) 116.45(2) 90.22(2) 667.0(3) 100/0 0.55(5) 2 2 Sample characterization 2 7bar H 7bar H C/5hC/5hC/5h 8.6038(8) 13.0198(9) 8.6057(7) 7.1860(4) 13.0284(6) – 7.1839(4) 8.593(2) – 12.963(2) 7.2223(6) 116.016(7) 90.62(2) – 116.003(6) 115.969(1) – 87.70(2) 722.6(2) 723.41(9) 0/100 723.91(8) 1/99 0/100 0.62(1) 0.99(2) 0.59(1) ◦ ◦ ◦ > >
Natural as well as commercially available minerals are often C/ 7bar H C/ 7bar H C/ C/ C/7bar H ◦ ◦ ◦ ◦ ◦ C/11d 8.5480(7) 13.0256(6) 7.1799(4) – 115.981(6) – 718.63(8) 15/85 0.58(1) impure. We checked for impurities using light microscopy ◦ of thin sections, which also allowed us to distinguish between different compositions and states of order. To determine the degree of Al,Si order, XRD patterns were obtained with a Philips X’Pert powder diffractometer, equipped with a primary Ge monochromator and with a Bruker D8 ADVANCE. For the XRD measurements, some sample material was crushed in an agate mortar to receive a “X-ray fine” grain size (approx. 1 μmto10μm), controlled by light microscopy. The powder was mixed with Si powder as internal standard to correct the zero shift of the X-ray diffractometers. Rietveld analysis of the XRD patterns were performed with the FULLPROF
software (Rodriguez-Carvajal 2005) to refine the lattice Overview of the used samples, their treatment, lattice parameters, Ab-Or content, and degree of order parameters. Atomic coordinates and lattice parameters for albite were taken from Wenk and Kroll (1984), for Numbers in brackets indicate uncertainty of the last digit Table 1 ID Sample name73125 Sanidine Albite129 Origin / precurser ID Sanidine Treatment Volkesfeld Cazadero 73 – a – b 850 c 8.5454(8) 13.0191(8) 8.132(2) 7.1827(9) – 12.783(2) 7.159(2) 94.29(2) 115.993(7) 116.59(2) – 87.69(2) 663.6(2) 718.27(9) 100/0 15/85 1.01(2) 0.62(1) 150 Or151 100 Prilep152 Microcline Microcline163 Or 100 Prilep Albite 151 129 125 – KCl/850 KCl/850 1060 8.580(2) 12.962(2) 7.2211(7) 90.65(2) 115.974(8) 87.69(2) 721.4(2) 5/95 0.99(1) 164 Albite165 Albite 125 125 1060 1060 166 Albite167 Albite 125 125 1060 1060 149 Or 100 73 KCl/850 456 M.P. Reitze et al. microcline from Strob (1983), and for sanidine from Kroll a grazing angle unit A513. We used a liquid nitrogen et al. (1986). Based on the obtained lattice parameters, Al,Si cooled MCT (mercury cadmium telluride) detector, and all site occupancies were derived from equations given by Kroll measurements took place at pressures of 2 hPa at room and Ribbe (1987)(Table1). temperature (ca. 23 ◦C). For the background spectrum we used a commercially available diffuse gold standard Sample treatment (INFRAGOLDTM). The powder samples were placed in an aluminum sample holder, 1cm in diameter and 1.5 For our investigation we needed alkali feldspars with mm in depth, and flattened with a spatula. Each grain endmember compositions, which possess a maximum size fraction was measured at two different geometries, spread of ordering states. Natural feldspars do not fulfill with an incident angle (i) of 20◦ and emergent angle (e) these requirements. Therefore, we performed annealing of 30◦ (i20◦ e30◦) and with an incident and emergent experiments to vary the state of order as well as alkali angle of 30◦ (i30◦ e30◦). For each sample and background exchange experiments to change compositions. Volkesfeld measurement, 512 single interferogram scans in double sanidine, ID 73, has been heated at 850 ◦C for 11 d (days), sided foreward-backward scan mode were summed up, and thereby slightly decreasing its state of order. Much larger the spectrum was calculated through Fourier transformation. changes were obtained for the Cazadero low albite. A In the wavelength region we measured, silicates show portion of ID 125 was filled in platinum tubes (ca. 1g per characteristic spectral features: (1) The Christiansen Feature tube) and sealed in evacuated silica glass tubes together (CF), which is a reflectance minimum at which the sample with gibbsite (=hydrargillite, Al(OH)3). The samples were is mostly transparent for the infrared radiation because there heated at 1060 ◦C to produce increasing states of disorder is a rapid change of the real part of the refractive index in (see Tab. 1 for details). H2O released by gibbsite acts as the sample (Conel 1969). This feature is diagnostic for the a flux that promotes disordering. The amount of gibbsite mineralogy (Conel 1969). (2) The Reststrahlen Bands (RB), was chosen so that the H2O pressure was about 7 bar. as a result of dominating surface scattering of photons. The samples were recovered after 8 d (ID 163), 15 d (ID RBs arise from the Si-O and Al-O stretching (e.g., Pieters 164), 18 d (ID 165), 47 d (ID 166) and finally after 92 d and Englert 1993; Iiishi et al. 1971). (3) The Transparency (ID 167). While Cazadero low albite is a pure Na-feldspar Feature (TF) results from increased volume scattering of the endmember, Volkesfeld sanidine and Prilep microcline infrared radiation and is only visible in samples with small contain some Ab component. They were thus treated by the grain sizes (< 25 μm) (Pieters and Englert 1993). alkali exchange method to produce K-feldspar endmembers Because peak fitting is complex (Malcherek et al. 1999), (Kroll et al. 1986). Charges of ID 73, ID 129 (63 μmto especially in powder reflectance spectra with broad and 125 μm) and ID 151 (100 μm to 125 μm) were heated at overlapping peaks, we decided to use the autocorrelation 850 ◦Cfor5hinanexcessofmoltenKClinacoveredPt method, which gives us a measure of the peak widths. This crucible in order to exchange Na ions by K ions. The weight method was used successfully to determine the degree of ratio sample/KCl was set at 1:4.5 (ID 73 → ID 129), 1:70 order in cordierite and Al,Ge ordering in Ba[Al2Ge2O8] (ID 129 → ID 150), and 1:5 (ID 151 → ID 152). Both small feldspar (e.g., Malcherek et al. 1995; Salje et al. 2000). and large weight ratios led to the desired results. After the Following this, we analyzed the feldspar infrared spectra not experiments, the samples were washed with distilled water, only in terms of band positions but also with autocorrelation filtered and then dried at 50 ◦C for several hours. The Or of specific band regions. For this we had to slightly modify contents of the exchange products were determined from the method given by Salje et al. (2000) and calculated the the position of the composition-sensitive (201)¯ reflection in following correlations. From the autocorrelation function X-ray powder diagrams (Kroll et al. 1986, their Table 9) we used the FWHM (Full Width at Half Maximum) of the resulting in Or99 and Or100, respectively (Table 1). complete autocorreleation function for two specific band regions from 8 μmto9μm and from 9 μmto10μm, Infrared and analysis procedure because of the remarkable changes in the spectra in these bands without baseline correction. Gaussian fits to the For the IR measurements, all samples were crushed in a central part of the autocorrelation function from 7.27 μm steel mortar and/or agate mortar and sieved to grain sizes to 12.11 μm were calculated with a constant offset around < 25 μm,25μmto63μm, 63 μm to 125 μm, 125 μm zero considering our spectral resolution (Offset values for − − to 250 μm, and grains greater than 250 μm. Mid-infrared fitting: -26 cm 1 and 26 cm 1. From this fit to the central measurements were performed with a Bruker 70v FTIR part of the autocorrelation function we used the width ω, −1 spectrometer with a spectral resolution of 4 cm using which corresponds with k2 in Salje et al. (2000). Mid-infrared spectroscopy of alkali feldspar samples for space application 457
Fig. 1 a Spectra of all samples with Ab100 with different degree of order and grain sizes between 63 μm and 125 μm(i20◦,e30◦). The gray region indicates the measurement range of MERTIS. The black lines indicate a feature shifting with t1 and Or content. The dashed lines indicate the borders of the region for autocorrelation analysis and the associated fit analyzes to determine the degree of order. The dashed-dotted lines indicates the bands for autocorrelation analyzes for the determination of the Or content. b The same as in a but for samples with Or100 to Or85
Results the XRD diagram. The calculated degree of order was also unchanged within the range of uncertainty (Table 1). X-ray diffraction The degree of order in the low albite sample ID 163 ◦ that had been heated for 8 days at 1060 C decreased from Heating of the Volkesfeld sanidine (ID 73) for eleven days t1 = 1to0.90;after15ditwasdownto0.86(ID ◦ at 850 C led to further disorder resulting in a final value 164). In the experiment that lasted 18 d (ID 165), the silica of t1 = 0.58 (ID 129). Heating of samples ID 73 and glass tube was bulging and was possibly broken, most likely ID 129 in molten KCl lead to an Or content of 100% in the because of a too high water vapor pressure. The crystals, samples ID 149 and ID 150. The degree of order remained however, were not molten. XRD analysis showed two albite unchanged. After ion exchange of the Prilep Microcline ID populations, 32% of the original low albite had not yet = 151 in molten KCl, the produced sample (ID 152) had an started to disorder, 68% were high albite with t1 Or content of 99% with no observable Ab component in 0.69. An average value of t1 = 0.79 was assigned to
Fig. 2 a Spectra of all samples with Ab100 with different degree of order and grain sizes smaller 25 μm(i20◦,e30◦). The gray region indicates the measurement range of MERTIS. The black lines indicate a feature shifting with t1 and Or content. The dashed lines indicate the borders of the region for autocorrelation analysis and the associated fit analyzes to determine the degree of order. The dashed-dotted lines indicates the bands for autocorrelation analyzes for the determination of the Or content. b The same as in a but for samples with Or100 to Or85 458 M.P. Reitze et al. the sample. The silica glass tube of sample ID 166 also et al. 1971), broaden and disappear with increasing disorder. strongly bulged but did not break and for this sample, the In the ordered species (ID 151, 152, 125), four clear bands degree of order was determined from XRD refinment to and one to two shoulders can be distinguished between −1 −1 t1 = 0.77. For sample ID 167, after opening the tube a 8 μm (1250 cm ) and 10.5 μm (952 cm ) in the region of small hole near the weld seam was noticed. At this position the RBs. In contrast, the spectra of disordered species show the tube wall was probably thinner because of welding only two or three broad peaks and two or three shoulders. and therefore a hole could form due to the water vapor Spectral differences resulting from the varying Or content pressure. The calculation of the degree of order gave a in the disordered samples (ID 73, 129, 149, 150) are only highly disordered t1 = 0.55. All samples, their precursor minor. The absolute intensities of the RB peaks change with material, treatment, refined XRD data, Ab-Or-content, as the degree of order as well. There are no observable RB well as degree of Al,Si order are listed in Table 1 (compare shifts resulting from Al,Si distribution changes. The mean section results). wavelength of the CFs is ca. 7.83 μm (1277 cm−1)forthe Or-rich samples (ID 73, 129, 149, 150, 151, 152) compared −1 Infrared to 7.72 μm (1295 cm )fortheAb100 samples (ID 125, 163, 164, 165, 166, 167). As an effect of small grain sizes, Our main goal was to investigate the spectral range spectral contrast and overall reflectance are reduced (Fig. 2). between 7 μm and 14 μm (marked by a gray background Thus, the peaks in the spectral region of the RB in samples in Figs. 1 and 2 /≈ 1429 cm−1 to 714 cm−1), with a high Al,Si order are weak but still distinguishable. which corresponds to the measurement range of the In the spectra of the more disordered species (ID 73, 129, MERTIS instrument. The overall reflectance of our samples 149, 150, 165, 166, 167), the small grain sizes also led to decreases with decreasing particle size because surface a decrease of the RB bands, so that the spectra are difficult scattering is decreased compared to volume scattering. The to distinguish. In addition, the TF is clearly visible in the autocorrelation function was used for the analysis of this spectra as a large broad peak between 11 μm and 12 μm spectral range (ranges of autocorrelation calculations: 8 μm (909 cm−1 and 833 cm−1). to 9 μm (1250 cm−1 to 1111 cm−1), 9 μmto10μm (1111 cm−1 to 1000 cm−1), and 7.27 μm to 12.11 μm (825 cm−1 to 1376 cm−1). Discussion Figure 1 shows the spectra of samples with a grain size of 63 μm to 125 μm, measured at i20◦ and e30◦ for samples General remarks with compositions Ab100 (Fig. 1a) and Or100 to Or85 (Fig. 1 b). The spectra change with the degree of Al,Si order. RB In the following, we discuss the infrared powder reflectance peaks, which are related to Si-O and Al-O stretching (Iiishi spectra results in terms of peak shift and the autocorrelation
Fig. 3 a Band position of the feature marked in Fig. 1 for grain sizes between 63 μmand 125 μm. Dashed-dotted line Or100 fit; dotted line Ab100 fit. b Same as a for figure but for grain sizes smaller 25 μm Mid-infrared spectroscopy of alkali feldspar samples for space application 459 analyzes for the three different grain size fractions. In addi- for the high albite sample it is shifted around = 0.12 μm tion, we debate implications that arise for remote sensing. towards shorter wavelengths (higher wavenumbers; = −1 IR absorption spectra show decreasing band intensities, 5cm ). In microcline with Or100 the feature is at 15.41 μm increasing linewidths, and shifts of bands with decreasing (649 cm−1) and shifts to 15.69 μm (637 cm−1) in highly Al,Si order (e.g., Zhang et al. 1997). Our powder reflectance disordered sanidine with Or100 (Fig. 3). Zhang et al. (1996) spectra exhibit the same changes. Zhang et al. (1997) reported an absorption feature in microcline also at around correlated peak positions and linewidths of absorption 15.43 μm (648 cm−1) and for highly disordered sanidine −1 −1 spectra above 16.6 μm (below 600 cm ) and in the far with Or100 at around 15.67 μm (638 cm ), which is infrared around 33 μm (300 cm−1) with Or content and in agreement with our reflectance data. These discussed degree of order of alkali feldspars. In contrast, our study features in the spectral range around 15.48 μm (646 cm−1) focused on the mid-IR, especially the wavelength region should be originating from bending of O − Si(Al) − O from 7 μmto14μm (1429 μm to 714 cm−1), which is the bonds (Zhang et al. 1996). The intensity of this bond measurement range of the MERTIS experiment (Hiesinger depends on the degree of order (Iiishi et al. 1971). Our et al. 2010; H. Hiesinger and co-workers, in preparation). results confirm these findings as seen in Fig. 3a. In addition, Following the classification of Iiishi et al. (1971), the the Or-Ab content also affects shape and position of the IR bands between 8.6 μm (1162 cm−1) and 17.2 μm analyzed feature in reflectance spectra. Therefore, we fitted (580 cm−1) should only depend on stretching and bending the wavelength of the 15.48 μm (646 cm−1) feature to of the bonds between Si or Al and O-ions. In contrast, we the degree of Al,Si order of the samples with Or100 and −1 found in our analysis, that this wavelength region is also Ab100, respectively. The 15.48 μm (646 cm ) feature in affected by the Or-content. This is probably associated with the spectra of samples with grain sizes smaller than 25 μm the change of the bond length
Table 2 Functions of the different fits calculated with the software gnuplot v.5.2
Correlation Figure Function Band position Or samples – t 63 μm to 125 μm3a−1.42 · λ + 22.91 100 1 Band position Ab samples – t 63 μm to 125 μm3a−2.02 · λ + 31.90 100 1 Band position Or samples – t < 25 μm3b−1.11 · λ + 18.00 100 1 Band position Ab100 samples – t1 < 25 μm3b−1.34 · λ + 21.38 Ratio of FWHM of autocorrelation functions () – Or-content 63 μm to 125 μm5a −321 · + 403 Ratio of FWHM of autocorrelation functions () – Or-content 25 μmto63μm5b −360 · + 454 Ratio of FWHM of autocorrelation functions () – Or-content < 25 μm5c−428 · + 568 ω – t all samples 63 μm to 125 μm6a−0.0005 · (26 − ω)2 + 1.02 Gauss 1 ω – t Ab 25 μmto63μm6b−0.032 · ω + 1.9 Gauss 1 100 ω – t Ab 25 μmto63μm6b−0.016 · ω + 1.5 Gauss 1 100 ω – t Or < 25 μm6c−0.12 · ω + 4.3 Gauss 1 100 ωGauss – t1 Ab100 < 25 μm6c−0.06 · ω + 2.8
λ in μm, is dimensionless, ω in cm−1 460 M.P. Reitze et al.
Fig. 4 Typical autocorrelation function (Auto.corr. is the value of the autocorrelation function) with marked FWHM and fit to the central peak between -26 cm−1 and 26 cm −1 (inset with ω of the Gaussian fit): ID 125 albite, spectral range 7.27 μmto 12.11 μm
wavelengths in the sample fraction with smaller grain sizes. Autocorrelation analysis This deviation is 0.01 μm larger than our estimated upper bound of the uncertainty of the feature wavelength of The calculations of the autocorrelation analyses were 0.05 μm. In particular, two microcline samples (ID 151, related to the degree of order and the Or content. Figure 4 152) and the albite sample (ID 125) show larger deviations shows a typical autocorrelation function. We tested several (125−25,MICRO,ALB ≈ 0.12 μm) from the large to the small different ways of combining autocorrelation analyses grain sizes. For all other samples, the mean deviation is at (linear, polynomial, and ratio correlations) and degree of the upper bound of uncertainty; therefore, the grain size- order as well as Or content. Ratios of different spectral dependent shift of the feature could be neglected. Based on regions are a common way in multi- or hyperspectral data our study we conclude, that with the feature shift analysis of analysis to identify distinct chemical compositions (e.g., the 15.28 μm (646 cm−1) feature, the degree of order of an Vincent and Thomson 1972; Kirland et al. 2002). In Fig. 5, alkali feldspar can be estimated if the chemical composition the ratio FWHM of the autocorrelation function of the is known. spectral range from 8 μmto9μm (1250 μm to 1111 cm−1)
Fig. 5 Correlation between Or content and , which is defined as the ratio of the FWHM of the autocorrelation function of the spectral range from 8 μmto 9 μmtothatofthespectral range from 9 μmto10μm(see dashed-dotted lines in Figs. 2 and 3) a Grain sizes of 63 μmto 125 μm; b Grain sizes from 25 μmto63μm, and c Grain sizes smaller than 25 μm. The yellow data point is sample ID 12 with Or62. Although it supports the linear trend between the endmembers it was not used for the fit because of contamination with quartz Mid-infrared spectroscopy of alkali feldspar samples for space application 461 to that of the spectral range from 9 μmto10μm (1111 μm of the smaller grain size fractions led to a split in the −1 to 1000 cm ) (denoted as thereafter) is plotted against correlation of ω and and degree of order for Or-rich samples =− · + the Or content for all analyzed samples and grain size and Ab-rich samples ( t1Or100 0.032 ω 1.9 and =− · + fractions. Samples with smaller Or contents have a larger t1Ab100 0.016 ω 1.5 for grain sizes from 25 μmto value than Or-rich samples. 63 μmand t1Or100 =−0.12 · ω + 4.3 and t1Ab100 = One measured sample with an intermediate chemical −0.06 · ω + 2.8 for the smallest grain size fraction). This composition between Or and Ab (Or62) suggest a linear implies that the grain size distribution has a large effect on relationship of value and Or content (yellow data point the form of the autocorrelation function in the regions of in Fig. 5). However, because of some accessory minerals the RB (Table 2). This is probably an effect of the reduced contained in the sample, it was not suitable for exact spectral contrast in combination with decreasing reflectance spectral analysis. To prevent uncertainties resulting of this in spectra of samples with smaller grain sizes compared to contamination we fitted our data with a linear function, those with larger grain sizes. The combination of the effects without using of the data of the sample with intermediate leads to a reduced FWHM of the peaks which is visible composition. The slope of the fitted straight line and its through the calculation of the autocorrelation function. intercept with the y-axis increase with decreasing grain Iiishi et al. (1971) stated that the mid-IR spectral region sizes, but the overall behavior is the same for all fractions of alkali feldspars is unaffected by alkali ions and only (Table 2). The respective fit functions are Or (mol%) = affected by the degree of order. In contrast, with the −321· +403 for samples with grain sizes between 63 μm autocorrelation analysis we can distinguish not only the and 125 μm, Or (mol%) =−360· +454 for samples with degree of order within infrared reflectance spectra between grain sizes between 25 μm and 36 μm, and Or (mol%) = 7 μm and 14 μm but also the Or content. Nevertheless, −428 · + 568 for the smallest grain size√ fraction. the reflectance spectral behavior is more complex, because In Fig. 6, the width ω = FWHM/ ln 4 of a Gaussian the analyzed grain size fractions have an influence on the fit (note: this FWHM is from the Gaussian fit and therefore RBs, which in turn have an influence on the autocorrelation something different as the FWHM from above) to the function. Therefore, knowing the grain sizes of the samples central peak from -26 cm−1 to 26 cm−1 on the x-coordinate is necessary for accurate determinations of Or content and of the autocorrelation function of the spectral band region degree of order from the spectra. Best results were obtained from 7.27 μm to 12.11 μm (1376 μm to 826 cm−1)is from samples with a grain size between 63 μm to 125 μm. plotted versus the degree of order. The black line in Fig. 6a is a quadratic fit to all the samples with a grain size ranging from 63 μm to 125 μm. The fit function to the dependence Conclusions of the width ω of the Gaussian fit curve on t1 is given by 2 t1 =−0.0005 · (26 − ω) + 1.02 (Fig. 6a). In Fig. 6b We investigated reflectance mid-IR powder spectra of and c, the results for the grain size fractions from 25 μmto endmember alkali feldspars with respect to their Ab-Or 63 μm and smaller than 25 μm are plotted. The analyses content and their Al,Si distribution which were derived from Fig. 6 Correlation between ω of Gaussian fit to the central part of the autocorrelation function of the wavelength region from 7.27 μm to 12.105 μm(see dashed lines in Figs. 2 and3) and the degree of order ( t1). a Grain size range from 63 μmto 125 μm. The black line is a quadratic fit to all samples. b Grain size range from 25 μmto 63 μm. Dashed-dotted line Or100 fit; Dashed line Ab100 fit, and c Same as b but with grain size smaller than 25 μm 462 M.P. Reitze et al.
XRD analysis. The results of this study showed that the Depending on the spectral deconvolution of data returned Al,Si degree of order as well as the Or content can be from MERTIS, it will be possible not only to determine determined by reflectance spectroscopy. Thus, these results the degree of order of an alkali feldspar, but also its are important and applicable for future space missions chemical composition. The data gained in this study carrying TIR spectrometers. as well as the calculated correlations are the basis for Mid-infrared spectra of alkali feldspars are strongly further investigations of the feldspar system, especially affected by the distribution of Al and Si on the non- for remote sensing of other planetary surfaces. Further equivalent tetrahedral sites of the frame-work structure and work on the alkali feldspar system will include samples are less affected by the amount of K and Na. Previous with intermediate chemical compositions plus mechanical work showed that the positions of specific absorption bands mixtures with the same bulk composition. are related to the Al,Si distribution (e.g., Martin 1970). On the basis of our study, we demonstrate that a specific Acknowledgements We are grateful to Stephan Klemme and Chris- feature found in all of our investigated spectra around tian Rengli for their help and permission to use their laboratory, 15.48 μm (646 cm−1), which is in the same spectral to Ludger Buxtrup for technical support with the furnaces, to Peter region as investigated by Martin (1970), is a measure Schmid-Beuermann for help with XRD at Institut f¨ur Mineralogie, and to Uta Rodehorst for advice and the possibility of using the lab- of the state of order. This feature shifts from 15.26 μm oratory at M¨unster Electrochemical Energy Technology. Constructive − − (655 cm 1) in pure low albite to 15.40 μm (649 cm 1)in comments by Charles Geiger, an anonymous reviewer and handling disordered high albite and from 15.41 μm (649 cm−1)in editor Anton Beran are gratefully acknowledged. This work is partly microcline to 15.69 μm (637 cm−1) in highly disordered supported by Deutsches Zentrum f¨ur Luft- und Raumfahrt funding 50 QW 1701 in the framework of the Bepi-Colombo mission. sanidine (Fig. 3). As different means to investigate the degree of order, Malcherek et al. (1995) introduced the autocorrelation analysis of absorption spectra. In Funding information Open Access funding provided by Projekt DEAL. contrast to the feature shift analysis discussed above, the autocorrelation analysis, which provides insights into the chemical composition as well as the degree of order, can be Compliance with Ethical Standards applied to the measurement range of MERTIS. We used the Conflict of interests The authors declare that they have no conflict of autocorrelation analysis on reflectance spectra for different interest. spectral regions. The ratio of FWHM of the autocorrelation function of 8 μmto9μm (1250 μm to 1111 cm−1)to the autocorrelation function of 9 μmto10μm (1111 μm Open Access This article is licensed under a Creative Commons to 1000 cm−1), called , is linearly dependent on the Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as Or content (Fig. 5). Iiishi et al. (1971) predicted that the long as you give appropriate credit to the original author(s) and the alkali ions do not affect the region of the RB in absorption source, provide a link to the Creative Commons licence, and indicate spectra of alkali feldspar. With our approach it is possible if changes were made. The images or other third party material in this to determine the Or content of (pure) alkali feldspars article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not from their reflectance spectra. Salje et al. (2000) suggested included in the article’s Creative Commons licence and your intended autocorrelation of the absorption spectra from 12.11 μm use is not permitted by statutory regulation or exceeds the permitted to 7.27 μm (825 cm−1 to 1376 cm−1) for cordierite and use, you will need to obtain permission directly from the copyright fitting of the central peak of the autocorrelation function holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/. with a Gaussian curve to investigate the degree of order. We used the same approach to analyze our powder reflectance spectra because MERTIS measures in the same spectral range and we are also observing spectral changes in these References wavelength region. For samples with grain sizes between Conel JE (1969) Infrared emissivities of silicates: Experimental 63 μm and 125 μm the correlation of t1 and the width results and a cloudy atmosphere model of spectral emission from ω of a Gaussian fit to the central part of the autocorrelation condensed particulate mediums. J Geophys Res 74:1614–1634. function is independent of the Or-content of the samples https://doi.org/10.1029/JB074i006p01614 and thus allows to identify the degree of order without Evans LG, Peplowski PN, Rhodes EA, Lawrence DJ, McCoy TJ, previous determination of the chemical composition. For Nittler LR, Solomon SC, Sprague AL, Stockstill-Cahill KR, Starr μ RD, Weider SZ, Boynton WV, Hamara DK, Goldsten JO (2012) samples with a grain size below 63 m, however, the Major-element abundances on the surface of Mercury: Results relation between ω and t1 becomes dependent on the Or from the MESSENGER gamma-ray spectrometer. J Geophys content. Res-Planet 117:E00L07. https://doi.org/10.1029/2012JE004178 Mid-infrared spectroscopy of alkali feldspar samples for space application 463
Hafner S, Laves F (1957) Ordnung/Unordnung und Ultrarotabsorption Pieters CM, Englert PAJ (1993) Topics in Remote Sensing 4. II. Variation der Lage und Intensitat¨ einiger Absorptionen Remote Geochemical Analysis: Elemental and Mineralogical von Feldspaten.¨ Zur Struktur von Orthoklas und Adular. Composition. Cambridge University Press Z Kristallogr 109:204–225. https://doi.org/10.1524/zkri.1957. Ribbe PH (1983) Feldspar mineralogy. De Gruyter. Berlin, Boston 109.1-6.204 Rodriguez-Carvajal J (2005) Fullprof Suite Laboratoire Leon´ Brillouin Hecker C, van der Mijde M, van der Meer FD (2010) Thermal (CEA-CNRS) CEA/Saclay, France infrared spectroscopy on feldspars - successes, limitations and Salje EKH, Wruck B, Kroll H (1985) Thermodynamics of sodium their implications for remote sensing. Earth-Sci Rev 103:60–70 feldspar II: Experimental results and numerical calculations. Phys Hiesinger H, Helbert J, Co-I Team MERTIS (2010) The Mercury Chem Minerals 12:99–107 radiometer and thermal infrared spectrometer (MERTIS) for the Salje EKH, Carpenter MA, Malcherek T, Ballaran TB (2000) bepicolombo mission. Planet Space Sci 58:144–165 Autocorrelation analysis of infrered spectra from minerals. Eur J Iiishi K, Tomisaka T, Katoˆ T, Umegaki Y (1971) The force field of K Mineral 12:503–519 feldspar. Z Kristallogr 134:213–229 Scambos TA, Smyth JR, McCormick TC (1987) Crystal-structure Jolliff BL, Wieczorek MA, Shearer CK, Neal CR (eds) (2006) New refinement of high sanidine from the upper mantle. Am Mineral Views of the Moon. Rev Mineral Geochem, vol 60. Mineral Soc 72(9-10):973–978 Am, Chantilly. 784 pp Solomon SC, Nittler LR, Anderson BJ (2018) Mercury: The View after Kirkland L, Herr K, Keim E, Adams P, Salisbury J, Hackwell J, MESSENGER. Cambridge University Press Treiman A (2002) First use of an airborne thermal infrared Sprague AL, Nash DB, Witteborn FC, Cruikshank DP (1997) hyperspectral scanner for compositional mapping. Remote Sens Mercury’s feldspar connection - mid-ir measurements suggest Environ 80(3):447–459 plagioclase. Adv Space Res 19:1507 Kroll H (1973) Estimation of the Al,Si distribution of feldspars from the lattice translations Tr[110] and Tr[110¯ ]. Contr Mineral Petrol Sprague AL, Donaldson Hanna KL, Kozlowski RWH, Helbert J, 39:141–156 Maturilli A, Warell JB, Hora JL (2009) Spectral emissivity Kroll H, Knitter R (1991) Al,Si exchange kinetics in sanidine and measurements of mercury’s surface indicate Mg- and Ca-rich anorthoclase and modeling of rock cooling paths. Am Mineral mineralogy, K-spar, Na-rich plagioclase, rutile, with possible 76:928–941 perovskite, and garnet. Planet Space Sci 57:364–383 Kroll H, Ribbe PH (1987) Determining (Al,Si) distribution and strain Strob WD (1983) Strukturverfeierung eines Tief-Mikroklins, Zusam- in alkali feldspars using lattice parameters and diffraction-peak menhange¨ zwischen